fix(allocator): BatchSandbox does not rebind a new Pod from Pool after allocated Pod is deleted#953
fix(allocator): BatchSandbox does not rebind a new Pod from Pool after allocated Pod is deleted#953longsuizhi wants to merge 3 commits into
Conversation
longsuizhi
requested review from
Generalwin,
Pangjiping,
Spground,
fengcone,
hittyt,
kevinlynx and
ninan-nn
as code owners
|
|
longsuizhi
changed the title
longsuizhi
changed the title
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e567cc496f
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d4b8e95d88
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
|
Changed directories: kubernetes. 📋 Recommended labels (based on changed files):
Other available labels:
💡 Tip: Use cc @longsuizhi |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5c9e7537fe
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
longsuizhi
force-pushed
the
fix/pool-rebind-deleted-pod
branch
2 times, most recently
from
53be88c to
56f2ed0
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 56f2ed04a4
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
longsuizhi
force-pushed
the
fix/pool-rebind-deleted-pod
branch
from
56f2ed0 to
ecdd9ea
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ecdd9eadef
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
longsuizhi
force-pushed
the
fix/pool-rebind-deleted-pod
branch
from
ecdd9ea to
20956e7
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 20956e73ce
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
当 Pool 中已分配给 BatchSandbox 的 Pod 被外部删除时，alloc-status 注解中仍保留 已删除 Pod 的名称，导致 supplement 计算为 0，无法触发重新分配。 本次修复在 getSandboxRequest 中增加了存活检测：将已删除的 Pod 从有效分配中 排除并加入 ToRelease 队列，使 supplement > 0 从而触发 Pool 重新分配新 Pod。
…ods from live allocation - P1: Released pods no longer count toward liveAllocated, preventing stale released entries from blocking subsequent re-allocations. - P2: Only Running+Ready pods are added to livePodSet, so terminal pods (Failed/Evicted) that still have their object present also trigger re-allocation.
longsuizhi
force-pushed
the
fix/pool-rebind-deleted-pod
branch
from
20956e7 to
46d6222
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 46d6222c71
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
2 participants
Fixes #954
问题描述
在 Pool 模式下，当已分配给 BatchSandbox 的 Pod 被外部删除（手动 delete、节点驱逐、OOM Kill 等）后，BatchSandbox 不会从 Pool 中重新获取新 Pod，导致 sandbox 永久不可用。
根因
allocator.go的getSandboxRequest通过alloc-status注解获取已分配 Pod 列表计算：Pod 被删除后注解未清理，
len(allocated)不变，supplement = 0，永远不会触发重新分配。修复方案
在
getSandboxRequest中增加存活检测：getSandboxRequestalloc-status中的 Pod 与实际存活 Pod，过滤出已删除的 PodToRelease队列清理分配记录测试结果
所有单测通过（含新增 3 个用例）：